Allocate as an oracle that knows the observed outcomes

Runs allocate() with degenerate predictive distributions that place all of their probability on the observed outcomes, giving the allocation against which a forecaster's allocation is scored. Because those distributions are point masses, the objective is piecewise constant and post_process() is always needed.

Usage

oracle_allocate(gpl_df, y, K, w = 1, ...)

Arguments

gpl_df

a gpl_df (see new_gpl_df()) or an allocated object, from which the loss parameters and weights are taken.

y

observed outcomes, one per target.

K

vector of budgets. Cannot be supplied via df.

w

numeric vector of costs per unit of resource allocated to each target.

...

further arguments passed to allocate().

Value

An allocated tibble; see allocate().

Examples

fc <- add_pdqr_funs(
  tibble::tibble(target_names = c("a", "b"), dist = "norm", mean = c(5, 8), sd = 1),
  types = c("p", "q")
)
a <- allocate(fc, K = 10)
oracle_allocate(a, y = c(4, 9), K = 10)$x
#> [[1]]
#>        a        b 
#> 3.076911 6.923089 
#>